Apply Machine Learning Algorithms to Predict At-Risk Students to Admission Period

Educational organizations have a high burden of treating students with poor academic success (At-risk students). Many methods are pursued to help this group of pupils, such as psychological, a proper schedule for the vulnerable pupil, recall, personal training, mock-tests, private education by direct studies, or the success centers. However, these approaches are not enough to deal with the dilemma, since other variables are impacting the success of the learner, which may be their family problems, their cognitive style, Their previous academic accomplishment, the cornerstone of the college level. This paper explores the effect of pre-college academic accomplishments on the success of college students in the computer program. The pilot study predicts student academic success based on their math (EmSAT Math), English (EmSAT English), and High School Average (High School Average) knowledge and skills acquired by students before college entry. The study covers 120 students. After cleaning up the available data, we are left with 112 instances used to train machine learning models to predict learner performance in the computer program. The qualified models achieved a high precision of 90.5 percent (Naive Bayes, Decision Tree, Random Forest), while the Logistic Regression achieved a forecast accuracy of 78.1 percent. The main symptom for the student's academic status was the score in English, followed by the score in Math. The research has shown that the high school score is not the secret to predicting students with academic status at their college stage.

I. Introduction

It is important to estimate the academic success of a student in any higher education. The reason behind this is that it is vital to determine which group of students can perform well in the semester final exam so that scholarships will be awarded, also to recognize those students who will fail in the semester assessments so that they could be given a sort of remediation. The academic achievement of students depends on several factors; some of them are past academic records, such as their performance in English, particularly if the English language is not the mother tongue and the educational medium uses the English language [1]. Learner's family history, mid-semester test results, etc. On the basis of these factors, the MLA classification models may indeed be configured to predict student outcomes. Various forms of MLA [2] have been used for this reason. This includes Naive Bayes, Decision Tree, Random Forest, Logistic Regression, to name a few. This research aims to collect data about new student’s pre-college academic performance, particularly their English-language score, which is essential for non-English speakers who study English curriculum, Math scores, and their high school scores. Data cleanest will be used to train four models, Naive Bayes, Decision Tree, Random Forest, and Logistic Regression, to identify key features that affect student performance at their next academic level. The association between the characteristics of the sample should be evaluated to identify the most significant factor affecting poor academic achievement (at-risk cases) and the main element contributing to outstanding academic success.